Upgrading Windows operating systems in an enterprise context can be a very expensive task. Not because the development of a new OS base install image is so complex, but the integration of all application packages in the image and the testing of those on the new platform are the main cost drivers.
Typically, the costs are split as follows:
- 33% developing the actual OS image
- 33% testing the applications on the new OS platform
- 17% project management and admin
- 17% actual roll out
To minimize Windows OS upgrade costs, but also to reduce costs for the running operation of a Windows based workplace, applications must be decoupled from the underlying operating system layer. When applications are decoupled (meaning, they are not directly installed on the Windows OS), the Windows base image is less complex (and therefore more easy to develop), and the integration testing of the all the applications on the new platform can be saved.
Now, decoupling is easy said, but hard achieved. New ways of how applications are provided to the user need to be explored. The most obvious way would be to drive applications to become HTML-5 based. Although this might be possible for applications which are newly introduced in the environment, we also need a solution for existing legacy programs. I see two main technologies which could be of use here:
- Application virtualization
- Application publishing
Application virtualization provides a sandbox around the application, so a new Windows platform does not interfere with the application context and the virtualized application in the sandbox can be deployed on the new OS without intensive testing. Further advantage of this technology is the possibility to manage the application from a central point and control application updates without the requirement to send out software update packages to thousends of clients.
Application publishing lets the software run on a terminal server. So again, the application context is independent from the workstation OS and can be upgraded centrally on the terminal server.
The more applications are provided with either option, the slimmer the OS base image can become and upgrades to new OS versions will be getting cheaper. This might be the first step to a new model of the client workplace where the actual client OS is not important any more, but just provides a runtime for required access software (like a browser or the Citrix receiver software).